To determine whether variables in the surgery resident selection process will predict attrition or performance during residency training. Retrospective study. A university residency program. A total of 111 categorical surgery residents matched during a 10-year period (1991-2000). Satisfactory outcome included successful completion of training and the American Board of Surgery examinations on the first attempt. Participants with a satisfactory outcome were stratified into good or marginal performance based on adverse actions during residency. Of 111 residents studied, 28 (25.2%) had an unsatisfactory outcome; attrition occurred in 25 (22.5%). Univariate analysis identified the following variables as predictors of unsatisfactory outcome: age at entry older than 29 years (P = .005), female sex (P = .02), courses repeated (P = .01), "C" grades on transcript (P = .01), no participation in team sports (P = .02), and lack of superlative comments in the dean's letter (P = .03). The following variables were retained in the multivariate model: age older than 29 years (odds ratio [OR], 0.11; 95% confidence interval [CI], 0.02-0.47; P = .003), summary comments in the dean's letter (OR, 4.57; 95% CI, 2.00-10.43; P < .001), participation in team sports (OR, 4.96; 95% CI, 1.36-18.05; P = .02), and merit scholarship in medical school (OR, 0.25; 95% CI, 0.08-0.78; P = .02). Attrition can be predicted from factors identified on residency applications, with nonacademic factors being more important. Among residents who completed the program, no predictors of performance were identified.